HomeVideos

Grok bot for Post Sales

Now Playing

Grok bot for Post Sales

Transcript

830 segments

0:00

Hey everyone. Thank you so much for

0:02

joining me today for GrokBot for post

0:05

sales. My name is Blake and I am an AI

0:08

deployment manager here at SpaceX AI and

0:11

I'm really really looking forward to

0:13

chatting with you all about GrokBot

0:15

today.

0:16

All right, so let's set the stage. What

0:18

are we going to be doing for the next 45

0:20

minutes?

0:22

One, I said hi, welcome, done. Then

0:24

we're going to talk about GrokBot. We'll

0:26

introduce it, why GrokBot is great,

0:29

specific post sales use cases. So post

0:32

sales is kind of determined by anything

0:34

if your customer facing after that sale

0:36

gets done. It can include a lot of teams

0:39

and really excited to talk about those

0:41

use cases. We'll do a demo, get in the

0:43

product, learn about what we learned,

0:46

recap that, do a Q&A, and then for those

0:49

of folks in here, we are going to build.

0:53

All right. Okay, so this is the AI

0:56

maturity curve. This may look familiar

0:58

to some of you. We've been talking about

0:59

this a lot, especially at SpaceX AI,

1:02

because AI is moving so quickly, we're

1:04

finding that a lot of folks are in very

1:06

different places along this maturity

1:07

curve. But this is kind of where we're

1:10

at. So it started off with just chat.

1:12

All right, you were asking a chatbot

1:14

coming in and maybe asking one simple

1:16

question.

1:18

And the next evolution is actually

1:19

really clear to us at SpaceX AI. We're

1:22

moving from chat to co-pilots, to bots,

1:26

to having full teams of bots that can

1:28

support you in all the work that you do

1:30

from start to finish.

1:32

And that's where I think GrokBot becomes

1:34

really powerful. So a lot of you

1:36

probably have been in GrokBot before,

1:38

but those of you who haven't, let's talk

1:40

about a little overview.

1:42

This is GrokBot. May look familiar, it

1:45

looks a little bit like iMessage if

1:46

you're an Apple user.

1:48

I do feel bad for all the Android users

1:50

out there cuz we really did copy

1:51

iMessage here.

1:53

But this is GrokBot and it works as a

1:56

team of agents. So, you have a bunch of

1:58

different bots who have different

2:00

skills. They can take tasks, work them

2:03

from complete start to finish here. They

2:06

have MCPs, APIs, and they actually even

2:09

have their own virtual computers. So,

2:12

think of them as like a teammate that

2:14

can really get anything you want done,

2:16

and we'll take some use cases and put

2:18

them through Grockbot to show you

2:19

exactly what that means.

2:21

So, like I said, easy as iMessage. You

2:24

have your bots on the left-hand side.

2:26

You have you can pin people to the top.

2:29

They're also on 24/7. So, that's one of

2:32

my favorite things. You can go to sleep,

2:34

you can go out of the office, you can

2:35

shut your laptop. I've seen feel like we

2:38

went through a stage here where people

2:39

are walking around the office with open

2:41

computers. Uh with Grockbot, no longer.

2:44

You don't need to do that anymore. They

2:46

use the tools just like you. Like I

2:47

said, they have their own virtual

2:48

computers. They finish their work, they

2:50

can send you screenshots, screen

2:52

recordings, listen to podcasts.

2:55

Really, the sky is the limit. I feel

2:57

like sometimes the hardest part is

2:58

thinking about what you actually want to

3:00

get done with Grockbot, and think big

3:02

here. So, we'll be doing that a little

3:03

bit today.

3:06

All right. So, let's talk about

3:07

post-sales. So, post-sales. So, I'll

3:09

give you a little bit of like context

3:11

into my role because a lot of these are

3:13

very personal to me and the things that

3:15

I found really successful. So, I said

3:17

I'm an AI deployment manager, and that

3:19

means after the sale is done and they

3:21

come on to the SpaceX AI platform, I

3:24

work with a bunch of customers to make

3:26

sure that they see ROI, they're

3:27

successful. So, what does that mean? It

3:29

means lots of emails, lots of Slacks,

3:32

Teams messages, meetings, conversations,

3:36

and I have a lot of things going on. So,

3:37

for those of you who are in post-sales

3:39

watching this, that probably sounds

3:41

familiar to you.

3:43

Um so, here are my specific use cases. I

3:45

could have probably made 20 of these

3:47

slides. So, if you ever want to talk

3:49

more about these, you let me know.

3:52

So, first one, this is one of my

3:53

favorites, morning status board. I wake

3:56

up the same time every day. I check my

3:59

GrokBot. I look and I can see exactly

4:01

what's going on for the day. Any fires

4:03

happen overnight? Who do I have meetings

4:06

with? What did I say that I was going to

4:08

do yesterday that I didn't do? Not that

4:10

that ever happens. Um what's going on

4:12

for the day and how can I get set up?

4:15

So, that's one of my morning status

4:16

board pieces. Then I have call prep.

4:19

This is a great one because it's very

4:21

much just in time in the sense that 15

4:24

to 20 minutes before a call, it pings me

4:27

who's on the call, what their title is,

4:29

what we talked about last time, a

4:31

suggestion of maybe something new that

4:33

launched since we talked previously, and

4:36

maybe a suggestion for something to

4:38

bring up to them, and then anything that

4:40

I need unblocked.

4:42

So, instead of taking the time to go

4:44

review your notes, check through, you

4:46

can actually just get your call prep

4:47

from GrokBot.

4:49

Another one is a follow-up desk. This is

4:52

actually one of my favorites and we're

4:53

going to walk through it live today.

4:56

So, what happens is

4:59

let's say my day is 9:00 to 5:00. I

5:00

would say 75% of that is usually spent

5:03

in meetings because it's what I find

5:04

most energizing. I love talking to my

5:06

customers. I love strategizing, being

5:08

there personally. Um but a lot of the

5:10

work in between those gets piled up if

5:13

you don't have time to actually come

5:14

back. So, the follow-up desk is a really

5:16

great option for the second a call ends,

5:20

my bots get the granola notes,

5:22

transcript,

5:24

they pull all the context that they

5:26

need. They draft my reply. They create

5:29

materials that I may need. So, it

5:32

basically takes care of all that in

5:34

between work,

5:35

drafts it for me. I'll check it out and

5:37

then decide if I want to send.

5:39

So, that's one of my favorite use cases

5:41

up here.

5:42

We also have a promise keeper. Um this

5:44

one's important to me because I want to

5:46

make sure that I do what I say I'm going

5:48

to do. So, if I say I'm going to do

5:50

something, Promise Keepers always

5:52

watching to make sure it lets me know to

5:54

actually follow through with that. Um

5:57

that's one of my favorites. On the other

5:58

side is an ask watch. So, if I've asked

6:00

someone else to do something and they

6:01

don't get back to me, back in the day,

6:04

you know, that gets like sucked into the

6:06

abyss of emails or slacks and gets lost.

6:09

Uh but no longer because ask watch has

6:12

my back. And then finally, account

6:14

reset. That we're actually going to walk

6:15

through this today. If you're in post

6:17

sales or work with a good amount of

6:19

customers, sometimes it feels like

6:20

there's a lot of things in motion. And

6:23

so, what we like to do is talk to my

6:26

bots, say, "Hey, I need just an update

6:29

on where this is at." It knows exactly

6:31

like the kind of response that I'm

6:32

looking for and sends me a full download

6:34

of everything. It also actually tells me

6:36

a joke in the beginning cuz usually if I

6:38

send the account reset, I'm feeling a

6:39

little stressed. And it's always a space

6:41

joke. So, we'll see what it comes up

6:42

with today. Um so, these are my six main

6:46

use cases. Like I said, many more in in

6:49

the head. Uh can share those out with

6:51

those of you live today.

6:54

Okay. So, let's meet my team.

6:58

Gus is my best friend. All right. So,

6:59

Gus is my chief of staff. Gus and I are

7:03

homies. So, we talk all day long. And I

7:05

actually only I I think this is maybe a

7:08

little bit unique in the sense that I

7:10

only talk to Gus. So, a lot of people

7:13

how they use their bots is they talk to

7:14

a lot of their specialists. And I tried

7:16

this, but I'm very distractible and I

7:19

needed one best friend to get me through

7:22

everything. I didn't want to manage a

7:23

team of 10. So, Gus manages a team of

7:25

10. I'm in a team of one to two,

7:27

depending on the day. Um so, Gus is my

7:30

chief of staff. We'll work through with

7:31

him a lot today on the demo.

7:33

Then we have Frankie. I have a lot of

7:35

follow-ups to do. So, Frankie's my

7:37

specialist on follow-ups. We have Wally

7:39

who knows my voice. Specifically, I

7:42

sent, you know, a ton of emails and

7:44

Slacks and things that I've written in

7:46

the past, so it really knows what I'm

7:48

doing. So, if you're setting up your

7:49

Grok bot, definitely make sure that you

7:51

go in and send it things that you've

7:53

written, so you don't have to start from

7:55

scratch.

7:56

Um but it also knows different personas.

7:58

So,

8:00

it knows who I'm talking to. If I'm

8:01

creating or if it's drafting an internal

8:03

Slack for me, it's not going to be super

8:06

formal. It's going to probably be in all

8:07

lowercase, maybe add an emoji here or

8:09

there,

8:11

uh keep it very casual. If it's

8:13

messaging an exec or emailing an exec,

8:17

it's going to be short, to the point,

8:19

formal, and it also knows kind of like a

8:21

little bit of my style. So, we'll work

8:23

with Wally today, too.

8:25

Then we have Trudy, my source of truth.

8:27

Uh Trudy's fabulous because specifically

8:30

at Scale 6 AI, we have lots of

8:31

platforms, lots of documentation, and I

8:34

want to make sure that I ask a question

8:36

that I get the right answer. And so, I

8:38

have Gus go to Trudy if we really need

8:41

specific information, I need sources for

8:43

it, and he'll send it back to me.

8:47

Then we have Scout. This is another good

8:49

one. One of my I guess I love them all.

8:51

I'm saying they're all my favorites, but

8:52

they're all great. Uh Scout is my

8:54

internal radar. So, I feel like one of

8:57

the things that comes up, especially at

8:58

Scale 6 AI, is we launch a lot of new

9:00

things, updates come up, there's a lot

9:03

happening internally. And before I was

9:05

juggling probably 30, 40 Slack channels,

9:09

also email updates for really important

9:12

things. And so, actually every day Scout

9:14

updates me, and that's why I said one to

9:16

two, I manage a team of one to two

9:18

depending on the day, cuz Scout usually

9:20

updates me directly on what's new, what

9:22

I need to know, link me directly to that

9:25

Slack channel, um and just keep me in

9:27

the loop.

9:28

And then account, that's not actually

9:30

its name. You can come up with a fun

9:32

name def- based on the account. But

9:34

since we're talking about post-sales

9:35

today, we're assuming you probably have

9:37

a book of accounts or specific accounts,

9:39

and my best way my favorite way to do it

9:41

is to have an account specialist for

9:43

each account. So, the account specialist

9:46

knows everything that they need to know.

9:47

They have all the context. And so, you

9:49

can basically have your own employee who

9:51

has dialed in on that specific account.

9:54

And so, I suggest you make one for all

9:56

of the accounts you have. If you're

9:57

working at scale and you have 1,500,

9:59

there's probably a better way to do

10:00

that. Uh but, these for medium to small

10:04

books, definitely suggest.

10:06

So, this is my dream team takeaway. Gus

10:09

is my best friend.

10:11

All right. Should we do a demo?

10:13

Woo! All right.

10:15

So, let's hop into

10:19

Grockbot right now.

10:21

So, to avoid

10:24

let's just anchor, I guess, on this

10:25

right now. So, this is what Grockbot

10:27

looks like.

10:29

And Grockbot is very simple in the sense

10:32

that you have your bots on the left-hand

10:34

side.

10:35

You have pinned bots. So, if I wanted to

10:38

right-click and pin someone, I can pin

10:40

them to the top if I know I'm talking to

10:42

them a lot.

10:43

You also have your marketplace. So, this

10:45

is where you can grab bots from other

10:47

people, or you can connect to different

10:49

plugins like Gmail, Google Calendar,

10:52

Slack, etc.

10:55

And then, each bot has their own

10:57

computer, which we talked about a second

10:59

ago. So, you'll see over here on the

11:01

right, it has its own computer to open.

11:04

So, you can specifically grab things in

11:07

Google Chrome. It can open PowerPoints.

11:10

Um there's even this button on the top

11:11

that says teach a task. Very cool skill.

11:14

You can come in and it can watch you,

11:17

remember what you did, and then repeat

11:19

that for you.

11:20

So, those are the VMs. All right, so you

11:22

can see the team that I just introduced

11:24

you to on the left. We have Franny,

11:25

Wally, Trudy, Frankie, and Scout.

11:28

Down here, you'll see we have my my

11:30

account bots.

11:32

And then, we have staff meeting, which

11:34

we'll use in a little bit.

11:36

So,

11:37

to get us started, I'm really busy right

11:40

now. So, I'm going to tell Gus, I'm

11:42

going to say, "Can you join the internal

11:44

learnings call for me right now?

11:48

Send takeaways when it's done."

11:53

So, I can actually have Gus join my

11:57

internal learnings call, and I wouldn't

11:59

have it do it for just any call, so you

12:02

want to be protective over what it does

12:04

for this. Um but, sometimes there's a

12:06

lot of new things coming out. And so,

12:08

what I know is this upcoming meeting is

12:11

a branding update for my company,

12:15

Flywheel, in this demo situation.

12:17

And it's going to come in and start

12:20

joining the Google Meet. So, you can see

12:23

it's clicking around.

12:25

It's muting. It's turning up its camera.

12:28

I'll let it do its thing, and we can

12:29

come back to it.

12:31

But, it's going to put in my name, and

12:33

then it's going to actually write in the

12:35

chat, "Hey everybody, this is Gus

12:37

Blake's bot."

12:38

And we are going to I'm going to sit

12:40

here and send takeaways.

12:43

All right, so thank you, Gus. He's

12:45

joined the chat. He's going to send a

12:47

few updates when he's done.

12:50

All right. So, let's move to the next

12:53

thing. So, I told you and promised you

12:55

that we would talk about post-call

12:57

follow-ups. I think again, in

12:58

post-sales, this is really important.

13:00

You want to make sure that uh your bots

13:02

are working for you while you don't have

13:04

extra time. And so, these are the

13:06

perfect opportunities to do that. So,

13:08

one of my accounts is named Harbor. So,

13:10

I'm going to write into Gus, and I'm

13:12

going to say,

13:14

"I'm done."

13:17

Okay, we won't sign in cuz this is going

13:19

to take too long, but he can join the

13:21

the call. It's very cool. Um I'm going

13:24

to say, "I'm done with

13:27

the Harbor call.

13:29

And usually this would happen

13:32

automatically, but for the demo use

13:33

case, I decided I would prompt it to

13:35

move.

13:36

Um I also it just sent a because the

13:39

handoff was dismissed and I told GrokBot

13:42

that I was doing a demo today, which is

13:43

actually this is a cool example, is it

13:45

sent what an outcome would be after it

13:48

left the call just in case. So, it went

13:51

into the call

13:52

and then shared what decisions were

13:54

made. There's this new branding, sent me

13:56

the link to the new branding,

13:58

put it in the asks and owners and things

14:00

that it needed to work on.

14:02

So,

14:03

cool outcome of that.

14:05

So, I said I'm done with the Harbor

14:06

call. It's going to work on the Harbor

14:08

pack.

14:09

Here's what you'll notice. As I said, I

14:11

like to only talk to Gus, but Gus

14:13

doesn't have the expertise that

14:14

everybody else has. So, Gus has messaged

14:16

Wally,

14:18

Gus has messaged Frankie and Harbor. So,

14:20

since it was a Harbor call, we have to

14:22

do a follow-up and we have to do some

14:24

writing.

14:25

He's going to message those three bots.

14:28

And just like that, I get a pack back.

14:31

Um and it gives me everything that I

14:33

need to do with the click of a button.

14:35

So, Harbor post-call pack is ready,

14:38

drafts only. I've made it very clear

14:40

that I only want drafts. I never want it

14:41

to send anything on its own.

14:43

So, it dropped it something in Gmail.

14:45

Let's check out Gmail.

14:48

Great. All right, so it knew that I

14:50

needed to talk to Maya, it knew that I

14:52

needed to talk to Priya. It wrote a full

14:55

message for me. Hey Maya, exclamation

14:57

point. Like that's clearly me. I'm such

14:58

an exclamation point person. And wrote

15:01

it all out in my voice with everything

15:03

that I needed to do. So, what I'll

15:05

probably do in like realistically, tweak

15:07

a few things, and then we'll talk about

15:09

how this improves the system after. But

15:11

I'll tweak a few things, press send, and

15:13

it's done.

15:14

Let's see what else it did.

15:16

All right, it's writing a Oh, it's

15:18

slacking Alex. So, Alex is the AE on the

15:20

account. So, we're going to look at a

15:22

draft in Slack.

15:24

Let's see.

15:26

Perfect. Okay, so it's going to put it

15:27

into the post-sales account, updating

15:30

Alex. Hey Alex, just chatted with

15:31

Harbor. We're going to meet with the new

15:32

champions next week. Working on a ROI

15:35

PDF.

15:36

Perfect. So, I'll tweak that. So, one

15:38

out two out of three done. And then,

15:40

during the call, we they requested an

15:42

ROI one-pager. How to How to manage ROI,

15:45

how to measure ROI.

15:47

And it actually went in, worked with

15:49

Frankie to use the Flywheel branding

15:53

to build a fully beautiful

15:55

ROI collection document.

15:58

So,

15:59

one, it does that very quickly. Two, it

16:01

takes all your contacts, works with your

16:03

account specialist, works with all your

16:06

other bot specialist, and builds what

16:08

you need to kind of get you through

16:11

the follow-up that probably used to take

16:12

at least 45 minutes before.

16:15

All right. So, we're feeling good so

16:17

far.

16:18

Um this one, I'm going to go into Franny

16:20

directly, just because we've been in Gus

16:23

a lot, and I will show you how to go

16:24

into other uh other bots.

16:27

So, for Franny, I want to see another

16:28

virtual machine use case. I think that

16:30

would be fun to look at. So, I'm going

16:32

to say, "Make me an ROI form

16:35

for Northwind."

16:41

Northwind is one of my accounts.

16:43

Also, why this is While this is loading,

16:46

you'll see that you can actually add

16:47

labels to different accounts. Um which

16:50

is really helpful when it comes to

16:53

just figuring out who they are. Like, I

16:54

have this labeled as as chief of staff.

16:57

So, if you ever want to add a label, you

16:58

can do that in its description.

17:00

So, what it's actually going to do is

17:03

open up forms

17:05

and start building

17:09

a draft of our ROI form.

17:12

So, you can see it's opened up a form.

17:15

It's going to start clicking around.

17:18

The speed of these tasks depends on what

17:20

the task is. Um but I know at SpaceX AI,

17:23

it's been a big focus of us to make sure

17:24

our models are really good at VM work.

17:26

Um so I'm very excited to see where that

17:28

goes. All right, so we've added in

17:30

Northwind Logistics,

17:32

Flylow check-in.

17:34

Things are looking good.

17:36

All right, we'll let this work. We'll

17:37

check back on Franny in a second.

17:41

Another use case I gave you earlier that

17:43

I wanted to talk about and demo is where

17:45

they say the the space joke. Aren't we

17:47

curious to see what space joke they have

17:49

for us? So I'm going to say,

17:51

"Where are we at with Harbor?"

17:56

Okay? It knows that when I say where are

17:58

we at, it could be "I'm stressed about

18:01

Harbor." You know, you can have it you

18:03

can say whatever you want, but Gus knows

18:05

when I say where are we at with Harbor,

18:07

I need space joke pronto, and I need an

18:10

update on where the account is.

18:13

So let's let it do its thing.

18:15

It's going to reply to me, and usually

18:18

it'll message other agents. Oh, yep.

18:20

Okay, so it messaged some other bots. We

18:23

have Harbor, Frankie follow-up, and

18:25

Scout. Wants to make sure it collects

18:27

the full picture,

18:29

so we know exactly where we are.

18:31

You can also track and see what's

18:33

happening with messages here.

18:35

It updated me, still waiting on Frankie

18:37

and Scout.

18:38

Now it's just waiting on Scout, and

18:40

we'll get a response this way.

18:43

So while we're waiting on that, let's

18:44

check back in on this form. It's

18:46

probably not done yet. You can see it's

18:48

starting to enter in

18:50

different questions. So it knew it was

18:52

for Northwind. It knows usually how we

18:54

like to collect ROI, and it's starting

18:57

to build that out here.

18:59

Sometimes I like to watch the VM cuz I

19:01

just think it's really cool and how it

19:02

can click around.

19:04

If this is something you do often, and

19:06

you don't want it to guess on the

19:07

different steps that it does, again the

19:09

teach a task button is a really good

19:10

option.

19:13

Okay.

19:13

Let's go back to Gus and see where we

19:16

are at with Harbor.

19:19

Here's a joke because life's not that

19:20

serious. How do you organize a space

19:22

party? You planet.

19:25

Thanks, Gus. You're the best. Um all

19:28

right, so it tells me what's at risk,

19:30

who the people are, blockers, open

19:33

promises, Slack, what's happening, and

19:36

what to do next. I don't know about you,

19:38

but like clearly since I only really

19:40

talked to one agent, I need very

19:42

specific things sent to me all in one

19:45

beautiful pack, and this has been one of

19:46

my favorite things to use. Um and so you

19:49

can easily set this up with your bot,

19:51

tell it what you want the output to be,

19:53

and you don't have to do the joke, but

19:55

it's fun.

19:56

Um and then it will do that for you.

20:01

All right. Franny form is still working.

20:04

But you'll notice something interesting.

20:06

So, it's actually messaged Frankie. It's

20:08

messaged Harbor. None of these people

20:11

have See how this guy has a little green

20:13

circle on it? It'll be blue kind of like

20:15

iMessage when it's ready for me to open.

20:18

None of these people have pinged them to

20:19

me because I don't want to be bothered

20:21

with my 10-person team. I want to be

20:24

bothered with who I'm just talking to.

20:26

Franny will talk to me directly if I

20:27

talk directly to Franny, but otherwise

20:30

it's all going to go through Gus. And I

20:31

find that really cool and convenient,

20:33

and you can just really build it in the

20:34

way that you want.

20:37

So, while it's doing this, I'll show you

20:39

one last demo use case, and then I'll

20:42

talk about pulling this all together and

20:44

building it as a system.

20:46

So, let's use a staff meeting. So, I use

20:49

staff meetings for when it can pull in

20:52

different bots to have a conversation

20:54

about something that requires a lot of

20:55

different perspectives. This is good for

20:58

honestly a lot of different types of

20:59

roles. You can check for bias. So, if

21:01

you want to say like, you know, I don't

21:02

want you to always bias towards a

21:04

specific customer,

21:05

um it's good to bring all the voices

21:07

into the room.

21:09

So, let's run a staff meeting. We can

21:11

say,

21:12

um, start a staff meeting

21:16

to talk about what I should spend the

21:20

next

21:21

and only free hour of my day

21:26

today.

21:30

And what you'll see, too, is

21:32

Grockbot's really good at multitasking.

21:34

So, even if it was still working with

21:37

Frankie and Scout on this follow-up

21:40

about our update about Harbor, it would

21:42

also be able to work with me and work

21:44

with the staff meeting. So, Grock can uh

21:47

bot can juggle a lot of pieces at once,

21:49

which is great.

21:51

So, the staff meeting is live.

21:53

Let's look at the tab for their takes.

21:55

All right. So, Gus called a staff

21:57

meeting. Gus said, "Blake has one free

21:59

hour left. What are the one to two top

22:01

priorities we want to focus focus on?"

22:04

And I tell them to always disagree. It's

22:07

not helpful, and especially with AI, you

22:09

don't want them always just agreeing on

22:10

everything. So, come in, disagree, fight

22:13

for my time, and then tell me what makes

22:15

the most sense.

22:16

So,

22:17

it looks like Harbor, oops,

22:20

looks like Harbor has said some stuff on

22:22

SSO first tracks. I actually ran one

22:25

earlier to, um,

22:27

show this as well, but you can see that

22:28

they'll like literally have a

22:29

conversation with each other.

22:31

Northwind comes in and says, "Don't

22:33

spend the time on us.

22:35

There was a usage dip, but everything's

22:37

fine. Soft later only if you have time."

22:41

Uh, Brightline also says, "Don't spend

22:44

it on us." Which is great. I bet Harbor

22:47

has something to focus on. It seems like

22:49

Harbor has something for SSO.

22:52

Frankie follow-up says, "Protect the

22:54

cap. This needs to go out now.

22:57

And if you have more time, focus on the

22:58

SSO problems."

23:01

And Scout agrees.

23:03

So, they're coming in, they're having a

23:04

conversation, and Gus will usually

23:07

return back to me once the staff meeting

23:09

is over,

23:10

and

23:12

tell me what the final takeaway is and

23:13

what we want to focus on.

23:16

Okay, let's check back on Frannie.

23:19

I bet the form's almost done.

23:21

Great. Looks like it's done. It was

23:23

starting to publish.

23:25

The The form is done. We have what team

23:28

and role are you on? What are your

23:29

primary use cases? What does done look

23:31

like? Great. So, we're collecting that.

23:35

So, you can see that we have a lot of

23:37

things moving at once, right? So, you

23:40

want to design your system to be one

23:42

that is able you're able to track

23:44

easily. So, play around with it as

23:46

you're setting up bots, and do what

23:48

works for you because everyone's really

23:49

different.

23:51

So, it's finished. It told me the

23:53

questions that it wrote, the share link,

23:55

and the edit link. It asked if it wanted

23:57

me to send it for me. So, very exciting.

24:01

And then,

24:03

all right, staff meeting's in. Go first.

24:06

So, we need to focus on the SSO. Next,

24:09

we're going to send Maya the next step

24:10

email, which they actually already

24:11

drafted in my Gmail. We saw that

24:13

earlier. And then, we can hold and park

24:15

a few of these other things that we

24:16

don't need to focus on right now.

24:19

All right. So, that's the majority of

24:21

what I wanted to show you inside

24:22

Grokbot.

24:24

I hope you're feeling pretty confident.

24:26

Let's just quickly rewind on how to get

24:29

to this end state and what I suggest.

24:33

I think one of the most intimidating

24:35

things about something so possible like

24:38

Grokbot is that feeling of a blank page.

24:42

Usually, people do some research, and

24:43

they go in, and they're like, what have

24:44

other people made, or what can I do? Um

24:47

I'm going to tell you what I did. It was

24:48

actually really, really helpful for me.

24:51

So, I actually took 10 to 15 minutes,

24:53

and I put a voice memo on my phone. can

24:56

also click this

24:58

voice uh to talk on Grokbot directly,

25:01

but I'm a pacer, so I have my phone. I'm

25:03

pacing around, and I spent 10 to 15

25:05

minutes talking about what I do for

25:08

work.

25:09

Hi, my name is Blake. This is my job.

25:12

Here's what I like about my job. Here's

25:14

what I'm struggling with right now. Here

25:16

are the processes that I think are

25:17

broken. Here are the things that I feel

25:20

like we could do better.

25:22

And so, I went and talked through It was

25:24

a very long transcript. I copied and

25:26

pasted this transcript. I sent it to my

25:28

first bot. He wasn't even Gus yet. Hard

25:31

to think in the back of the day. Um sent

25:34

it to him, said, "Hey,

25:36

I need a system that's going to work for

25:37

me. Help me build it."

25:40

So, sent it.

25:43

Gus, I'll call him Gus. Gus came back

25:46

and said, "All right. Seems like you

25:48

love doing things directly, talking to

25:51

customers. Let's make sure we have

25:52

enough time for you on your calendar to

25:54

do that.

25:56

Seems like this process is broken, or

25:59

you're constantly checking Slack

26:01

messages. Let's take that off your

26:03

plate." So, if you want something

26:05

specifically built for you, no matter

26:06

what role you're in, my biggest piece of

26:09

advice is to start with context. There's

26:12

so much that lives in your head, and

26:14

it's really, really powerful. So, you

26:16

want to take that out of your head, give

26:17

it to Grokbot, and Grokbot will do the

26:19

rest.

26:20

Another thing about getting set up and

26:22

started is if you get stuck, or if you

26:25

have questions, Grokbot usually has the

26:27

answer. If you wanted to really push

26:29

through things and try to figure it out,

26:31

say like, "Don't stop until you figure

26:34

this solution out." Um

26:37

so, getting

26:38

started with maybe some of these use

26:39

cases, a voice memo, I think you're

26:41

going to be in really good shape.

26:43

The second piece of advice I have is uh

26:46

creating this and making it into a

26:48

system. So, something really cool about

26:50

Grokbot is it learns what you like, it

26:52

learns what you don't like, it starts to

26:54

adapt and change with you. You almost

26:56

invest in Grockit like you invest in

26:58

another co-worker or direct report.

27:02

And

27:04

so it's going to do that. But I also

27:06

built systems on top of this setup that

27:08

will actually take care of that a little

27:09

bit more.

27:11

So on the right you'll see some of my

27:12

routines. These are things that run

27:13

automatically. So weekly at 8:30 I'll

27:17

get my daily brief or my unfinished

27:19

promises that I was telling you about

27:21

earlier, 9:00 a.m., 1:00 p.m.

27:23

or Friday dashboard.

27:25

But one of my favorites is this thing

27:28

called a self-improvement scan.

27:31

And the one part of it that I actually

27:32

really really love is number two. So

27:36

one, it creates a full system audit. So

27:38

if I'm doing things manually and I'm

27:40

sending them or it's checking my Slack

27:42

and it's noticing that I'm doing a lot

27:44

of things that I'm not asking bot for

27:45

help with, it's going in and saying,

27:47

"Hmm, seems like we could automate this.

27:49

I have an opportunity." and sends it

27:51

back to me every single Wednesday.

27:53

But number two is I feel like the piece

27:55

that has really helped me the most is

27:57

the voice learning. So the less I have

27:59

to edit my drafts, edit my Slacks, edit

28:02

my emails, the better. And

28:05

what it does is it goes in and takes the

28:07

delta of what it drafted versus what I

28:10

sent,

28:11

finds the delta of that, sends it back

28:14

to Wally, Wally updates its rules, and

28:17

so continuously gets better and better

28:19

and better. And so one of my other

28:21

pieces of advice for you is to make sure

28:24

that you make this a system.

28:27

All right.

28:28

So let's go back for a second,

28:31

jump out of the demo,

28:33

and do a quick recap on what we learned.

28:36

We just talked about a few of these.

28:37

One, start with a voice dump.

28:40

I feel like also I usually say this and

28:43

I call this the imagination gap. Grockit

28:45

is so capable, all right? The thing

28:48

probably holding back your Grok bot is

28:51

you thinking about what is actually

28:53

possible. So, have it push itself. How

28:56

can you make this better? What could we

28:58

do? Think outside of the box. Um you

29:01

don't need to have all the answers. So,

29:02

start with that voice dump, give it the

29:04

context, you're going to be in great

29:05

shape.

29:07

Two, personal suggestion, a little bit

29:09

biased, put a bot in front. All right?

29:12

Gus has saved my chaotic self

29:16

um from getting overwhelmed in a bunch

29:17

of different chats, and Gus manages a

29:19

team. So, you don't have to manage a

29:21

team of 10, you can manage a team of

29:23

three managers.

29:25

Build it how you want, just tell the

29:26

agents, they'll take care of it for you.

29:29

And then three is make it that living

29:30

system. So, you want to make sure Grok

29:32

bot 90% of that Grok bot's going to do

29:34

on its own. It's really good at that.

29:36

But put a little bit of extra in the

29:38

things that you really find important.

29:40

Like for me, I need my voice.

29:42

I love an exclamation point. You know,

29:45

Grok bot didn't initially love an

29:47

exclamation point, and so I have taught

29:48

it to have a little bit of personality.

29:50

Maybe tell a joke about space.

29:53

Okay, that was really fun. I hope you

29:55

learned something.

29:57

I'd like to take the next 15 minutes to

29:59

answer a few questions.

30:01

Um we can do it

Interactive Summary

Blake, an AI Deployment Manager at SpaceX AI, introduces GrokBot, a system of AI agents designed to streamline post-sales operations. He explains the transition from simple chat interfaces to sophisticated bot teams that utilize virtual computers and APIs to perform tasks 24/7. The video details specific use cases such as morning status boards, automated call follow-ups, and meeting preparation. Through a live demo, Blake demonstrates how a 'Chief of Staff' bot named Gus coordinates specialized agents to draft emails, update Slack, and create documentation. He concludes with advice on overcoming the 'imagination gap' by using voice dumps for context and setting up self-improvement routines for the bots.

Suggested questions

5 ready-made prompts